CVE-2026-42271

HIGH8.8

CISA KEV

Published 2026-05-08 · Updated 2026-10-26 · Source security-advisories@github.com

Summary

CVE-2026-42271 is a high-severity (CVSS 8.8) vulnerability in BerriAI LiteLLM. CISA added it to the Known Exploited Vulnerabilities catalog on June 8, 2026, which means exploitation in the wild has been confirmed; US federal agencies had to remediate it by June 22, 2026. CISA does not currently list it as used in ransomware campaigns. BerriAI has 3 vulnerabilities in the KEV catalog, 0 of them linked to ransomware.

All 3 known-exploited BerriAI vulnerabilities →

Description

LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. From version 1.74.2 to before version 1.83.7, two endpoints used to preview an MCP server before saving it — POST /mcp-rest/test/connection and POST /mcp-rest/test/tools/list — accepted a full server configuration in the request body, including the command, args, and env fields used by the stdio transport. When called with a stdio configuration, the endpoints attempted to connect, which spawned the supplied command as a subprocess on the proxy host with the privileges of the proxy process. The endpoints were gated only by a valid proxy API key, with no role check. Any authenticated user — including holders of low-privilege internal-user keys — could therefore run arbitrary commands on the host. This issue has been patched in version 1.83.7.

CVSS vector: CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H

CWE-77CWE-78CWE-78

CISA Known-Exploited Vulnerability

Product: BerriAI — LiteLLM

Name: BerriAI LiteLLM Command Injection Vulnerability

Date added: 2026-06-08 · Due: 2026-06-22

Required action: Apply mitigations per vendor instructions, follow applicable BOD 22-01 guidance for cloud services, or discontinue use of the product if mitigations are unavailable.

References

View on NVD

More exploited BerriAI vulnerabilities

Most recently added to CISA KEV

CVE-2026-59822
LiteLLM — BerriAI LiteLLM Improper Authentication Vulnerability
added 2026-09-02
CVE-2026-42208
LiteLLM — BerriAI LiteLLM SQL Injection Vulnerability
added 2026-05-08

View all →

Related CVEs

Same severity, most recent

CVE-2026-68933
Chartbrew is an open-source web application that can connect directly to databases and APIs and use the data to create charts. Prior to 5.2.3, public dashboard passwords were stored in plaintext, accepted through a URL query parameter, and checked without effective attempt throttling. The vulnerable surface is public password-protected dashboard storage and validation routes. An unauthenticated attacker targets a public password-protected dashboard and repeatedly submits candidate passwords or obtains password-bearing server or proxy logs. Unlimited online guesses enable brute force, while cleartext storage and URL transport expose the secret in application storage and server or proxy access logs. The target must be a public dashboard with password protection enabled. Even failed attempts reveal chart count and configuration without the password; a recovered password grants access to the protected dashboard and its chart data. This vulnerability is fixed in 5.2.3.
2026-11-06 · score 7.5
CVE-2026-50240
A flaw was found in OpenShift Pipelines-as-Code (PAC). PAC does not verify whether the creator of a Repository custom resource has permission to create PipelineRuns. A namespace user with only Repository and Secret create/delete permissions can cause PAC to execute attacker-controlled PipelineRuns as the auto-provisioned pipeline ServiceAccount, which carries ClusterRole/edit by default. This results in namespace-scoped privilege escalation, enabling Secret read, resource creation/modification/deletion, and potential supply chain injection in CI/CD namespaces.
2026-11-06 · score 8.8
CVE-2026-123791
A vulnerability was detected in OpenBMB Internet of Agents up to 9d78e88e47160eb42ee1ee8538c05e1c2c47877a. This affects the function _primary_field_query of the file common/utils/milvus_utils.py of the component Agent Registry. The manipulation of the argument key results in improper neutralization of special elements in data query logic. The attack may be launched remotely. The exploit is now public and may be used. This product operates on a rolling release basis, ensuring continuous delivery. Consequently, there are no version details for either affected or updated releases. The project was informed of the problem early through an issue report but has not responded yet.
2026-11-06 · score 7.3
CVE-2026-85736
Chartbrew is an open-source web application that can connect directly to databases and APIs and use the data to create charts. Prior to 5.2.3, connection creation or update stored attacker-chosen absolute SSL and SSH file paths, and later connection cleanup passed those stored paths to fs.unlink without a path restriction. The vulnerable surface is data-connection certificate and SSH-key cleanup when a connection is removed. A user with createOwn and deleteOwn permissions on their own team stores a target path in a connection and then deletes the connection to trigger cleanup; default self-signup grants teamOwner these permissions. The cleanup routine treats the stored SSL or SSH path as a managed certificate path and unlinks the referenced file; duplicateConnection provides a second advisory-identified trigger. The attacker must be able to create or update and then delete a connection on their own team. The default container runs Chartbrew as uid 0, while deployments with a less-privileged process are limited by that process's deletion permissions. The attacker can delete files such as /code/.env or application source, causing integrity loss and service disruption. This issue is fixed in version 5.2.3.
2026-11-06 · score 8.1
CVE-2026-85726
Chartbrew is an open-source web application that can connect directly to databases and APIs and use the data to create charts. Prior to 5.2.3, a date-typed MongoDB variable placed inside a quoted query placeholder was inserted as an unescaped string, unlike the safely escaped or coerced string, number, and Boolean branches. The vulnerable surface is mongoDB-backed public, shareable, or embedded charts processed by date-variable substitution. An unauthenticated user supplies a crafted URL filter to a public, shareable, or embedded MongoDB chart whose query places a date-typed variable in a quoted literal. The crafted date value breaks out of the quoted literal and changes query logic; the chart's returned row count becomes a Boolean oracle for character-by-character field extraction. The chart must be externally reachable and use the vulnerable date-in-quoted-placeholder branch. Extraction is read-only and limited to one queried collection per request, and it requires neither an account nor a known field value. The blind oracle can disclose any field in the queried collection, including secrets the chart does not display. This issue is fixed in version 5.2.3.
2026-11-06 · score 7.5